Simulation Neuronaler Netze auf Massiv Parallelen Rechnern

Zell, Andreas; Mache, Niels; Hüttel, Markus; Vogt, Michael

doi:10.1007/978-3-642-78486-6_88

Cited by 14 publications

(15 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…volume and surface area of the grain phase per unit reference volume) from a given training set of preclassified data (Tewari, 1997; Wei et al ., 1998; Mattfeldt et al ., 1999). From the rich choice of available neural network types, we decided to use classical multilayer feedforward networks with backpropagation (MLFF‐BP networks) and networks based on learning vector quantization (LVQ networks) (Zell, 1994; Kohonen, 1997) (Fig. 3).…”

Section: Methodsmentioning

confidence: 99%

“…In these studies initialization of LVQ networks was deterministic, hence there was no need for repetition of runs. For the studies with MLFF‐BP networks, we used the resilient backpropagation algorithm (Riedmiller & Braun, 1992, 1993), implemented within the Stuttgart Neural Network Simulator (Zell, 1994). For LVQ studies we used the LVQPAK software (Kohonen, 1997).…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Classification of spatial textures in benign and cancerous glandular tissues by stereology and stochastic geometry using artificial neural networks

Mattfeldt¹,

Gottfried²,

Schmidt³

et al. 2000

Journal of Microscopy

View full text Add to dashboard Cite

Stereology and stochastic geometry can be used as auxiliary tools for diagnostic purposes in tumour pathology. Whether ®rst-order parameters or stochastic-geometric functions are more important for the classi®cation of the texture of biological tissues is not known. In the present study, volume and surface area per unit reference volume, the pair correlation function and the centred quadratic contact density function of epithelium were estimated in three case series of benign and malignant lesions of glandular tissues. The information provided by the latter functions was summarized by the total absolute areas between the estimated curves and their horizontal reference lines. These areas are considered as indicators of deviation of the tissue texture from a completely uncorrelated volume process and from the Boolean model with convex grains, respectively. We used both areas and the ®rst-order parameters for the classi®cation of cases using arti®cial neural networks (ANNs). Learning vector quantization and multilayer feedforward networks with backpropagation were applied as neural paradigms. Applications included distinction between mastopathy and mammary cancer (40 cases), between benign prostatic hyperplasia and prostatic cancer (70 cases) and between chronic pancreatitis and pancreatic cancer (60 cases). The same data sets were also classi®ed with linear discriminant analysis. The stereological estimates in combination with ANNs or discriminant analysis provided high accuracy in the classi®cation of individual cases. The question of which category of estimator is the most informative cannot be answered globally, but must be explored empirically for each speci®c data set. Using learning vector quantization, better results could often be obtained than by multilayer feedforward networks with backpropagation.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Classification of spatial textures in benign and cancerous glandular tissues by stereology and stochastic geometry using artificial neural networks

Mattfeldt¹,

Gottfried²,

Schmidt³

et al. 2000

Journal of Microscopy

View full text Add to dashboard Cite

show abstract

“…From the many available ANN types, a classical multilayer feedforward network with backpropagation (MLFF‐BP) and learning vector quantization (LVQ) [ 25, 26] were used, the latter in two software packages, the SNNS and LVQPAK implementations. The MLFF‐BP networks are the most popular ANNs by far, already implemented in various academic and commercial software packages.…”

Section: Methodsmentioning

confidence: 99%

Prediction of prostatic cancer progression after radical prostatectomy using artificial neural networks: a feasibility study

Mattfeldt¹,

Kestler

Hautmann

et al. 1999

BJU International

View full text Add to dashboard Cite

Objective To report a methodological feasibility study in correctly classify new cases was tested using the 'leaveone-out' technique. a small series of patients with node-negative organconfined prostatic cancer, using artificial neural netResults Progression was predicted correctly in 85% of newly presented cases from the three routine histopaworks to predict tumour progression after radical prostatectomy and thus help to identify high-risk thological variables alone. On the basis of the four morphometric variables alone progression was prepatients who would benefit from adjuvant treatment. Patients and methods A group of 20 patients with dicted correctly in 93% of cases. The use of all seven variables as input data only slightly improved the pT2N0 prostatic cancer and postoperative tumour progression was compared with a control group of 20 quality of prediction. The best results were obtained by the LVQ networks and LDA, followed by MLFF-BP patients with no progression, matched for age, duration of follow-up and preoperative serum prostatenetworks. Conclusions In this methodological feasibility study, the specific antigen level. Histopathological data were obtained from the radical prostatectomy specimens, progression of pT2N0 prostatic cancer after radical prostatectomy could be predicted with good accuracy, i.e. the Gleason score, World Health Organisation (WHO) grade and maximum diameter of the tumour sensitivity and specificity from routine variables or morphometric texture variables using artificial neural transects. The volume and surface area of the epithelial tumour component and of the lumina of the neoplastic networks. These results suggest that this approach should be assessed in a prospective study with more glands per unit tissue volume were estimated by morphometric methods. To predict recurrence, multicases. Keywords Artificial intelligence, information science, layer feedforward networks with backpropagation (MLFF-BP), two implementations of learning vector neural networks, pathology, prostatic neoplasms, urology quantization (LVQ), and linear discriminant analysis (LDA) were applied. The ability of these models to unnecessarily to the risk of side-eCects, which may be

show abstract

“…Neural networks have been employed for many classification tasks in Bioinformatics and Computational Biology. EvoDNN is a framework that employs an Evolutionary Algorithm (EA) to evolve the weights, biases and AFs of a deep heterogeneous feed-forward neural network (FNN) [87] (A neural network (NN) that consists of many hidden layers, often more than five, is called a deep neural network). EvoDNN extends our earlier framework EvoNN [70], which evolved a simple, single hidden layer FNN.…”

Section: Motivationmentioning

confidence: 99%

“…Finally, gradient descent methods struggle to train deep neural networks [29], caused by gradient vanishing and explosion [40], saddle points, and some other forces, which not only affect deep FNNs [87], but also recurrent networks [64]. However, the depth of the network has no impact on neural networks trained by EAs, because EAs are a zero-order, population-based method.…”

Section: Motivationmentioning

confidence: 99%

EvoDNN - An Evolutionary Deep Neural Network with Heterogeneous Activation Functions

Cui

Shabash

Wiese

2019

2019 IEEE Congress on Evolutionary Computation (CEC)

View full text Add to dashboard Cite

Many problems in Computational Biology and Bioinformatics involve classification, such as the classification of cell samples into malignant (cancer) or benign (normal). For such tasks, we propose EvoDNN, an evolutionary deep neural network that employs an evolutionary algorithm to evolve deep heterogeneous feed-forward neural networks. While the majority of current feed-forward neural networks employ user defined homogeneous activation functions, EvoDNN creates heterogeneous multi-layer networks where each neuron's activation function is not statically defined by the user, but dynamically optimized during evolution. The main advantage offered by EvoDNN lies in that the activation functions do not need to be differentiable. This feature gives users a great degree of flexibility over which activation functions EvoDNN can utilize. This thesis demonstrates how EvoDNN can simultaneously optimize each neuron's weight, bias, and activation function, and empirically shows a superior performance compared to feed-forward neural networks trained with backpropagation method, random forest method, and our earlier approach EvoNN which employed a single hidden layer.

show abstract

Simulation Neuronaler Netze auf Massiv Parallelen Rechnern

Cited by 14 publications

References 2 publications

Classification of spatial textures in benign and cancerous glandular tissues by stereology and stochastic geometry using artificial neural networks

Classification of spatial textures in benign and cancerous glandular tissues by stereology and stochastic geometry using artificial neural networks

Prediction of prostatic cancer progression after radical prostatectomy using artificial neural networks: a feasibility study

EvoDNN - An Evolutionary Deep Neural Network with Heterogeneous Activation Functions

Contact Info

Product

Resources

About